In the new servers we've recently bought I've gone for the Dell RD1000 drives for backups which work pretty well, looking to move to a d2d2t solution eventually or maybe just scattered NAS boxes - unsure as yet, but I haven't had a problem with HDD backups yet (touch wood)!
Tapes expensive and struggling to cope with capacity needed now. When tape autoloader died I went to NAS box then copy to 2TB e-sata drive for off -site.
When I did a refresh of our server equipment I bought a tape library to do backups (I've since added in a NAS/ISCSI box to do D2D2T). After 18 months of use, the tape drive failed, for no reason - was working one day, powered the box off and on correctly (doing some major power system rearrangement), drive stopped working. Cost for a replacement drive - £1700! Slightly stupidly I hadn't looked at the length of warranty, but as it's enterprise equipment, I'd assumed 3 years, but it was only 1.
We're currently doing D2D (to separate building) which is OK (some stability issues, but that may be due to cryptd on the NAS or the HDDs in the NAS), I'm looking to use Ahsay "offsite" backup software in future to backup to this NAS and then replicate to truely offsite (co-lo at LA datacentre). We can then provide offsite backup to primaries as well :)
Speaking from first hand experience of a total disaster recovery (gas explosion next door wiped all the disks (including the backup ones! in D2D2T setup).
the disk backups were gone (all disks shocked beyond there spec and scrambled)
The tapes were ok BUT (we use E 12.5), we learnt the hard way that you can't restore from the copy. You need to restore the restor files back to the disks then restore from there!
We are moving away toatlly from BE and onto VEEM.
Just our experience, but we actually had the 'acid test'! All servers and disk backup trashed in a few milliseconds!!
What did come out of this is the number of people that have never actually done a full DR and so do nont know that there backup solution does NOT work.
What ever you choose, please make sure that it actually does work fully. We were down for a week restoring, when we thought we would be just a day or two.
I'll provide my two cents. Assuming you don't have the luxury of offsite d2d, and if you really want to go down that route, then ideally you need to look at an integrated solution that uses both media. Initial outlay may be more, but you need to see the wider longer term cost savings and not the immediate capital expenditure. I've designed and implemented a few backup solutions over the years and a combination of both is suited to most situations, obviously discounting disk to disk datacentre/enterprise solutions where offsite snaps are taken at a location significantly distanced from the primary site (unlike Salan's unfortunate incident)
Every environment is different and in some cases can be really limited in terms of design, but your solution is primarily determined by cost, existing infrastructure placement, setup and capability and not to mention your DR/BCP requirements. However many people don't really investigate this properly, and it's something that needs some degree of planning (data sizing, throughput, location, mechanism (NDAS, SAN, Tape, Fiber, Network etc), backup windows, retention periods, archiving policies and so on)
I've recently proposed a small "upgrade" solution for a corporate virtual environment consisting of a Scalar i500 LTO4 solution and a 4TB dual GbE connected DXi4520 De-Dupe unit for d2d capability. Because there was no fiber connectivity I had to use network which of course throughput can be a factor as tape drives suffer from “shoe-shining” (This is the stop-start element during tape backup). The introduction of disk improved that performance, if only to the level allowed by the network. I used the dedupe unit between the master server and tape library as a buffer and the used that to stream data at a nominal level across the network, it also reduces tape media needed and can be controlled by the same backup software, in this instance Veritas Netbackup.
The disk i then used for incremental daily backups and then the tape unit used as a second process for full backups that require a longer retention period. It also allowed backup to tape at more convenient periods which could also be sent offsite (less frequently that previous, and reducing tape requirements significantly so lesser ongoing cost) The disk backup was set as the primary target with duplicate copies of the data moved to tape as a secondary process outside of the backup window.